A HYBRID SEMANTIC SIMILARITY FEATURE-BASED TO SUPPORT …eprints.uthm.edu.my/id/eprint/10184/1/Nurul_Aswa_Omar.pdf · 2018-06-20 · A HYBRID SEMANTIC SIMILARITY FEATURE-BASED TO

A HYBRID SEMANTIC SIMILARITY

FEATURE-BASED TO SUPPORT MULTIPLE

ONTOLOGIES

NURUL ASWA BINTI OMAR

UNIVERSITI TUN HUSSEIN ONN MALAYSIA

A HYBRID SEMANTIC SIMILARITY FEATURE-BASED TO SUPPORT

MULTIPLE ONTOLOGIES

NURUL ASWA BINTI OMAR

A thesis submitted in

fulfillment of the requirement for the award of the

Doctor of Philosophy

Faculty of Computer Science and Information Technology

Universiti Tun Hussein Onn Malaysia

AUGUST, 2017

iii

DE In the name of Allah, The Most Beneficent, The Most Merciful.

Thank you Allah for giving me such wonderful people.

Deep appreciation to my beloved husband,

Beni Widarman Bin Yus Kelana

My children,

Muhammad Faris Aiman

Nur Fatihah Aleeya

My parents and father-in-law,

Omar Othman; Shakinah Yaakob; Yus Kelana; Noraini

and my siblings (Nurul Atiqah, Muhammad Syafiq, Nurul Athirah, Muhammad

Syamil, Muhammad Syarafi).

Thank you for your prayers, understanding, caring, compromising and everything.

DICATION

iv

ACKNOWLEDGEMENT

Firstly, I would like to express my thanks to Assoc. Prof. Dr. Shahreen Binti Kasim,

and Assoc. Prof. Dr. Mohd Farhan Bin Md Fudzee for their willingness to accept me

as a Ph.D. student. Their guidance, support, determination, encouragement,

understanding and patience along this journey is greatly appreciated.

I am also grateful to the Ministry of Higher Education (MOHE) for

sponsoring me under the Scheme of Academic Training for Public Higher Education

Institution (SLAI) during my studies.

I am also greatly indebted to the Faculty of Computer Science and

Information Technology (FSKTM) and the Centre for Graduate Studies (CGS) of

Universiti Tun Hussein Onn Malaysia (UTHM) for providing good facilities and an

inspiring environment for me to complete this study comfortably.

I am dedicating special thanks to my friends, especially the postgraduate

members (Dr. Isredza, Dr. Norhanifah, Norhanani, Norlida, Mahfuzah, Yana

Mazwin and Dr. Yusliza) for their help, motivation, encouragement and knowledge

sharing throughout this journey. Many thanks to Dr. Arfian Ismail, Dr. Sazali Khalid,

Dr. Riswan Effendi for helping in completing my thesis journey.

Last but not least, I would like to thank my husband, my son, my parents, my

father-in-law and my siblings for their prayers, support, understanding, compromise

and everything.

v

ABSTRACT

Semantic similarity between concepts, words, and terms is of great importance in

many applications dealing with textual data, such as Natural Language Processing

(NLP). Semantic similarity is defined as the closeness of two concepts, based on the

likeliness of their meaning. It is also more ontology-based, due to their efficiency,

scalability, lack of constraints and the availability of large ontologies. However,

ontology-based semantic similarity is hampered by the fact that it depends on the

overall scope and detail of the background ontology. Coupled with the fact that only

one ontology is exploited, this leads to insufficient knowledge, missing terms and

inaccuracy. This limitation can be overcome by exploiting multiple ontologies.

Semantic similarity with multiple ontologies potentially leads to better accuracy

because it is able to calculate the similarity of these missing terms from the

combination of multiple knowledge sources. This research was conducted for

developing the taxonomy of semantic similarity that contributes to understanding the

current approaches, issues and data involved. This research aims to propose and

evaluate ontological features for semantic similarity with multiple ontologies.

Additionally, this research aims to develop and evaluate a feature-based mechanism

(Hyb-TvX) to measure semantic similarity with multiple ontologies which can

improve the accuracy of the similarity. This research used two benchmark datasets of

biomedical concepts from Perdesen and Hliaoutakis. Similarity value, correlation and

p-value were also used in the evaluation of the relationship between the concept pair

of multiple ontologies. The findings indicate that the use of a semantic relationship

of concepts (hypernym, hyponym, sister term and meronym) can improve the

baseline method up to 75%. Besides that, the Hyb-TvX mechanism produces the

highest correlation value compared to the other two methods, that is 0.759 and the

result correlation is significant. Finally, the ability to discover similarity concepts

with multiple ontologies could be also exploited in other domains besides

biomedicine as future research.

vi

ABSTRAK

Persamaan semantik antara konsep, perkataan dan terma adalah penting dalam

pelbagai aplikasi yang berkaitan dengan data teks seperti pemprosesan bahasa tabii.

Persamaan semantik merupakan satu pendekatan bagi mengenalpasti persamaan

konsep melalui perbandingan makna. Persamaan semantik lebih cenderung

menggunakan ontologi berdasarkan kecekapan, berskala, kurang kekangan serta

mempunyai ontologi yang besar. Walaubagaimanapun, penggunaan ontologi di

dalam persamaan semantik masih dibatasi oleh kebergantungan ontologi yang

terperinci dan hanya satu ontologi diterokai yang menyebabkan ketidakcukupan

pengetahuan, kehilangan terma dan ketidaktepatan persamaan. Permasalahan ini

boleh diatasi dengan mengeksplotasi kepelbagaian ontologi. Persamaan semantik

dari pelbagai ontologi berpontensi meningkatkan ketepatan persamaan dengan

pengiraan persamaan dalam situasi kehilangan terma serta gabungan pelbagai sumber

ontologi. Kajian ini dijalankan untuk membangunkan taksonomi persamaan semantik

dalam memahami pendekatan semasa, isu dan data yang terlibat. Kajian ini bertujuan

mencadangkan dan menilai ciri-ciri ontologi untuk persamaan semantik dari pelbagai

ontologi. Disamping itu, kajian ini juga bertujuan untuk membangunkan dan menilai

mekanisme berasaskan ciri-ciri (Hyb-TvX) bagi pengukuran persamaan semantik

pelbagai ontologi dalam meningkatkan nilai ketepatan persamaan. Kajian ini telah

menggunakan dua penanda aras set data bioperubatan konsep yang terdiri daripada

Perdesen dan Hliaoutakis. Nilai persamaan, kolerasi dan nilai p juga digunakan bagi

melihat perhubungan dan kepentingan hubungan antara konsep dari pelbagai

ontologi. Dapatan kajian menunjukkan penggunaan perhubungan semantik konsep

(hypernym, hyponym, sister term dan meronym) telah meningkatkan kolerasi

sebanyak 75%. Selain itu, kaedah pengukuran Hyb-TvX menghasilkan nilai kolerasi

yang tinggi berbanding dua kaedah sebelum ini iaitu 0.759 dengan keputusan

kolerasi signifikan. Akhir sekali, penyelidikan persamaan pelbagai ontologi boleh

dieksploitasi dalam bidang selain bioperubatan di masa depan.

vii

TABLE OF CONTENTS

DECLARATION ii

DEDICATION iii

ACKNOWLEDGEMENT iv

ABSTRACT v

ABSTRAK vi

TABLE OF CONTENTS vii

LIST OF TABLES xii

LIST OF FIGURES xvi

LIST OF ALGORITHMS xix

LIST OF SYMBOLS AND ABBREVIATIONS xx

LIST OF APPENDICES xxv

LIST OF PUBLICATIONS xxvi

CHAPTER 1 INTRODUCTION 1

1.1 Overview 1

1.2 Problem statement 3

1.3 Research objectives 5

1.4 Research questions 5

1.5 Scope and research significance 6

1.6 Organization of the thesis 9

viii

CHAPTER 2 LITERATURE REVIEW 10

2.1 Introduction 10

2.2 Similarity 11

2.3 Words similarity 12

2.4 Semantic similarity 16

2.5 Knowledge-based 19

2.5.1 Ontology-based 19

2.5.2 General purpose ontologies used with

semantic similarity approaches 24

2.5.3 Domain specific ontologies used with

semantic similarity approaches 27

2.6 Classification of semantic similarity approaches

according to ontology 33

2.7 Single ontology 37

2.7.1 Structure-based approach 37

2.7.2 Information content-based approach 40

2.7.3 Feature-based approach 42

2.7.4 Hybrid-based approach 43

2.8 Multiple ontologies 43

2.8.1 Matching method 44

2.8.2 Ontological feature matching for

multiple ontologies 45

2.8.3 Semantic similarity measure for

multiple ontologies 47

2.9 Biomedical domain as a case study 56

2.10 Discussion of similarity 59

2.11 Chapter summary 61

CHAPTER 3 RESEARCH METHODOLOGY 63

3.1 Introduction 63

3.2 Structure of semantic similarity 63

3.3 Research framework 64

3.4 Data sources and preparation 68

3.4.1 WordNet 71

ix

3.4.2 MeSH 72

3.5 Instrumentation and result analysis 72

3.5.1 Hardware and software requirements 74

3.5.2 Testing and analysis 74

3.6 Parameter 78

3.6.1 The proposed parameter 79

3.6.2 The classical probabilistic theory 92


CHAPTER 4 THE ONTOLOGICAL FEATURES

ALGORITHM IN SEMANTIC SIMILARITY FOR

MULTIPLE ONTOLOGIES 95

4.1 Introduction 95

4.2 Ontological features algorithm 97

4.2.1 Matching matrix semantic relationship 97

4.2.2 Semantic overlapping between

subsumers 99

4.2.3 Selecting the LCS 101

4.3 Result and discussion 103

4.3.1 Ontological features evaluation result 103

4.3.2 Matching method results 106

4.3.3 Correlation between similarity

measurement and method for multiple

ontologies 109


CHAPTER 5 Hyb-TvX: A HYBRID SEMANTIC SIMILARITY

FEATURE-BASED MEASUREMENT FOR

MULTIPLE ONTOLOGIES 113

5.1 Introduction 113

5.2 Hyb-TvX: A hybrid semantic similarity feature-

based measurement 116

5.2.1 TvX-1: similarity measurement level 1 122

5.2.2 TvX-2: similarity measurement level 2 125

5.3 Result and discussion 128

x

5.3.1 Experimental results 128


CHAPTER 6 CONCLUSION AND FUTURE WORKS 135

6.1 Introduction 135

6.2 Research summary 135

6.3 To develop the taxonomy of semantic similarity

that contributes understanding towards current

approaches, issues and data related to the topic 137

6.3.1 What is the appropriate approach for

semantic similarity with multiple

ontologies in order to understand the

issues and data involved? 137

6.4 To propose and evaluate an ontological features

algorithm in semantic similarity for multiple

ontologies in terms of features matching

accuracy 138

6.4.1 Does the feature-based measurement

used is appropriate in measuring the

ontological features algorithm

proposed? 138

6.4.2 Does the proposed ontological features

algorithm increase the accuracy of

feature matching? 139

6.4.3 Is there any relationship between

dependent variable (human scored) and

independent variables (similarity values

for each method)? 140

6.5 To develop and evaluate feature-based

mechanism to measure semantic similarity of

multiple ontologies to increase similarity

accuracy 141

xi

6.5.1 How does the proposed parameters

support in improving the accuracy of

similarity for multiple ontologies? 141

6.5.2 Does it a significant relationship

between the Hyb-TvX methods with

human scored? 143

6.6 Contribution of the research 143

6.7 Future research 144

REFERENCES 145

APPENDIX 157

xii

LIST OF TABLES

2.1 Symbol and meaning in relation to similarity 12

2.2 Words similarity approach 15

2.3 Fields and applications that use the semantic

similarity approach 17

2.4 Advantages and disadvantages based on the

background information in the semantic

similarity approach 19

2.5 Classification approach according to the

categories of ontology 38

2.6 Advantages and disadvantages of the

Tversky method 42

2.7 Types of ontological features in each

semantic similarity method 45

2.8 Descriptions of two concepts from WordNet

and MeSH 49

2.9 Method feature-based approach for multiple

ontologies 56

2.10 Summary of datasets for single ontology

used in previous work 58

2.11 Summary of datasets for multiple ontologies

that used in previous work 58

3.1 Set of 36 medical term pairs with averaged

experts similarity scores (Hliaoutakis et al.,

2006) 68

3.2 Set of 30 medical term pairs with averaged

physician and coder ratings scores (Pedersen

et al., 2007) 70

xiii

3.3 The level of strength correlation coefficient 77

3.4 Example of results human similarity and

proposed method 77

3.5 Calculate the correlation coefficient (r) table 77

3.6 Result implementation parameters for

situation 1 80


situation 2 81


situation 3 82


situation 4 83


situation 5 84


situation 6 85


situation 7 86


situation 8 87

4.1 Comparison previous method and OntF 102

4.2 Similarity result of the proposed method

(OntF) using four similarity feature-based

measurements in Pedersen benchmark

datasets based on physician ratings 104



measurements in Pedersen benchmark

datasets based on coder ratings 105



measurements in Hlioutakis benchmark

datasets based on expert ratings 106

xiv

4.5 Similarity results of multiple ontologies. The

bold column represents the proposed method

(OntF) 108

4.6 Correlation values based on four similarity

feature-based measurements using different

methods. The row in bold shows the results

of the proposed method (OntF) 110

4.7 Improvements that the OntF achieved over

the average of the three other methods using

the Sánchez measure based on Physician,

Coder and Expert ratings 111



the Jaccard measure based on Physician,




the Dice measure based on Physician, Coder

and Expert ratings 112



the Ochiai measure based on Physician,


5.1 Example for concepts extracting a set of

token 118

5.2 Example for concepts with tokenization of

words 118

5.3 Example for concepts and synonym 119

5.4 Examples of concepts 122

5.5 Example for concepts and synonym 124

5.6 Example of concepts and features 126

5.7 Comparison similarity of the proposed

method with the X-similarity and Rodriguez

xv

and the Egenhofer method using Pedersen

benchmarks datasets 129

5.8 Comparison similarity of the proposed

method with the X-similarity and Rodriguez

and Egenhofer method using Hlioutakis

benchmarks datasets 130

5.9 Comparison similarity of proposed method

with physician, coder and experts ratings

(averaged) 131

5.10 Correlation of similarity method on feature-

based approach for multiple ontologies

according to the WordNet and MeSH dataset 133

5.11 p-value of similarity method on feature-

based approach for multiple ontologies 133

6.1 Investigated the proposed parameters (α) in

eight situations involve condition |comp B|,

|comp A|, differentiation complement and

intersection value 143

xvi

LIST OF FIGURES

1.1 Example of terminological matching

between subsumer of antibiotic (WordNet)

and antibacterial agent (MeSH) 4

1.2 Comparison of the number of articles

published every year and the number of

citations per year

(http://apps.webofknowledge.com) 8

2.1 Content structure 10

2.2 The computer ontology 20

2.3: Semantic similarity in single ontology for

male and female WordNet ontology 22

2.4 Semantic similarity in multiple ontologies

for intracranial hemorrhage (in Snomed-CT)

and brain neoplasms (in MeSH) 23

2.5 The snapshot of lexical database for english

(WordNet) 25

2.6 The snapshot of WordNet content for renal

failure 25

2.7 The snapshot of the Cyc web page. 26

2.8 Snapshot of the UMLS web page. 28

2.9 The snapshot of Medical Subject Headings

(MeSH) 29

2.10 The snapshot of MeSH content for kidney

disease 29

2.11 The snapshot of Snomed-CT web pages 30

2.12 An example of a biological process tree for

biological_process (GO: 0008150). It is the

xvii

parent while biological regulation

(GO:0065007), localisation (GO:00051179),

phosphorus utilisation (GO:0006794), and

signaling (GO:00023052) are its children. 32

2.13 The snapshot web page of SDTS 33

2.14 Semantic similarity for single ontology 34

2.15 Semantic similarity for multiple ontologies 35

2.16 Single ontology 37

2.17 The concept similarity between C1 and C2 40

2.18 WordNet taxonomy, showing most-specific

subsumers of nickel and dime and of nickel

and credit card. Solid lines represent is-a

links; dashed lines indicate that some

intervening nodes have been omitted 41

2.19 Ontologies 1O and 3O 54

2.20 The process to develop hybrid semantic

similarity feature-based for multiple

ontologies 62

3.1 Structure of semantic similarity method 64

3.2 Research framework 65

3.3 Research structure 66

3.4 Example of a taxonomical tree in WordNet 73

3.5 Example of a taxonomical tree MeSH 74

3.6 |comp B| is larger than |comp A| 87

3.7 |comp B| is smaller than |comp A| 88

3.8 Results for condition 1 in situations 1, 2, 3

and 4 90

3.9 Results for condition 2 in situations 5, 6, 7

and 8 91

4.1 Ontological modelling of the concept lupus

in MeSH 98

4.2 MeSH ontology for myocardium concept 100

5.1 The flow process of Hyb-TvX 115

xviii

5.2 Venn diagram for intersection 116

5.3 Venn diagram for union 117

5.4 Venn diagram for complement 117

5.5 Intersection of 1C and 2C 118

5.6 Maximum tokenization of 1C and 2C 119

5.7 Intersection synonym of 1C and 2C 120

5.8 Union synonym of 1C and 2C 120

5.9 Venn diagram for complement A 121

5.10 Venn diagram for complement B 121

xix

LIST OF ALGORITHMS

4.1 The proposed ontological features algorithm

(OntF) 101

4.2 Features algorithm from Sánchez et al.,

(2012) 102

4.3 Features algorithm from Batet et al., (2013) 102

5.1 The process Hyb-TvX algorithm 128

xx

LIST OF SYMBOLS AND ABBREVIATIONS

- For all, any, each, every of element.

- Is element of

- Selection of

- Greater than or equal to

A - Set A

B - Set B

o - Objects

- Cross

R - Real number

= - Similar to

x - Concept x

y - Concept y

- Intersection

* - Multiplication

1C - First concept

2C - Second concept

3C - Third concept

1N - 1N are the number of is-a links from A

2N - 2N are the number of is-a links from B

3N - 3N are the number of is-a links from C

LCS - Lower common subsume

W(c) - W(c) is the set of words (nouns) in the corpus

c / C - Concept (c) or concept (C)

- Total

xxi

N - The total number of word (noun) tokens in the corpus

baS , - Similarity between concept a and b

21,CCSim - Similarity concept 1 and concept 2

baSim , - Similarity concept a and concept b

O - Ontology

O1 - First ontology

O2 - Second ontology

Ri - Type (i) of semantic relationship for first ontology

Rj - Type (j) of semantic relationship for second ontology

X - X correspondences to sets of a

Y - Y correspondences to sets of b

YX / BA - Set X union set Y or Set A union set B

|X - Y| - The relative complement of Y in X

XY - The relative complement of X in Y

BA - Set A union set B

1 > 0 - 1 must be more the zero

- Parameter for complement

padepth - Depth of ontology for concept a

qbdepth - Depth of ontology for concept b

qp ba , - Parameter for complement a and b

0,, nuw www - Weighting parameter must be same or more than zero

wS - Similarity word matching

uS - Similarity feature matching

nS / odneighborhoS - Similarity neighborhood

qp

p baS , - Similarity parts for concept a and b

qp

f baS , - Similarity function for concept a and b

qp

u baS , - Similarity attribute or concept a and b

synsetS - Similarity synset or synonym

ndescriptioS - Similarity description

hypoTotal - All hyponym

xxii

iihypo SOtotal - All hyponym in ontology i for subsumer i

(r) - Pearson correlation coefficient

n - Number of word pairs

ii yx - Total multiplication of human judgments ( ix ) and

iy is the corresponding ith element in the list of

similarity value

ix - Total value human judgment or physician judgment

iy - Total value similarity based similarity measurement

method

MR - Matrix relationship

MS - Matrix subsume

Ri (Si) - The row represent the subsumer of relationship

Rj (Sj) - The column represents the subsumer of relationship

Hyb-TvX - Proposed method (A Hybrid Semantic Similarity

Feature-based Measurement)

TvX-1 - Similarity Measurement level 1

TvX-2 - Similarity Measurement level 2

(Int) - Intersection

(Un)/U - Union

(comp) - Complement

(max) - Take the maximum value of the words

Ax - x element of set A

Bx - x element of set B

comp A/ 'A - Complement for set A

comp B/ 'B - Complement for set B

Ux - x element of set Union

Ax - x not element of set A

21 ,CCInt - Intersection for concept 1 and concept 2

21 ,max CC - Maximum for concept 1 and concept 2

BAInt , - Intersection set A and set B

BAUn , - Union set A and set B

xxiii

21,CCSc - Similarity concept for concept 1 and concept 2

21,CCSs - Similarity synonym for concept 1 and concept 2

21,CCS f - Similarity features for concept 1 and concept 2

cS - Similarity concept

sS - Similarity synonym

fS - Similarity features

fsc SSS ,,max - Identify value maximum between fsc SSS ,,

aw - The proposed parameter for |comp A|

bw - The proposed parameter for |comp B|

GB - Gigabyte

RAM - Random Access Memory

Ghz - Gigahertz

PHP - Hypertext preprocessor

GIS - Geographic information systems

STS - Semantic textual similarity

WordNet - Ontology WordNet

Snomed-CT - Systemized Nomenclature of Medicine Clinical Term

MeSH - Medical Subject Heading

GO - Gene ontology

IC - Information content

UMLS - Unified Medical Language System

ICD - International Clasification Disease

S - Synonym

SENSUS - Ontology SENSUS

Cyc KB - Cyc knowledge base

KB - Knowledge base

CPT - Current procedural terminology

ICD-10-CM - International Classification of Diseases, Tenth

Revision, Clinical Modification

LOINC - Logical Observation Identifiers Names and Codes

xxiv

NLM - National Library of Medicine

MH - Mesh Heading

STDS - Spatial Data Transfer Standard

LCA - Lower common ancestor

STS - Semantic textual similarity

RM - Root Matching

TM - Terminological Matching

TM - Terminological Subsumption

SS - Semantic Subsumption

LCS - Lower common subsume

xxv

LIST OF APPENDICES

A1 List of total concept datasets WordNet for

Hliaoutakis et al., (2006) benchmark 157

A2 List of total concept datasets MeSH for

Hliaoutakis et al., (2006) benchmark 159

A3 List of total concept datasets WordNet for

Pedersen et al., (2007) benchmark 161

A4 List of total concept datasets MeSH for

Pedersen et al., (2007) benchmark 162

xxvi

LIST OF PUBLICATIONS

Journal:

(i) Nurul Aswa Omar, Shahreen Kasim, Mohd Farhan Md Fudzee (2016) (in

press) “A Review of Semantic Similarity Approach for Multiple

Ontologies.” International Journal of Information and Decision Sciences.

(Scopus Indexed)

(ii)

Nurul Aswa Bt Omar, Shahreen Kasim, Mohd Farhan Md Fudzee (2016)

(in press) “Extended TvX: A New Method Feature Based Semantic

Similarity for Multiple Ontology.” Journal of Telecommunication,

Electronic and Computer Engineering (JTEC) Special Issue (Scopus

Indexed).

1CHAPTER 1

INTRODUCTION

1.1 Overview

The birth of the internet has led to an abundance of textual information in the World

Wide Web. By using different language terminologies, people can access

information from various sources in several formats (mostly text). However,

resources are hard to manage due to the lack of textual understanding capabilities of

computerized systems. The estimation of the semantic similarity between concepts,

words, and terms is of great importance in many applications dealing with textual

data, such as natural language processing (NLP) (Patwardhan & Pedersen, 2006),

knowledge acquisition (Sánchez & Batet, 2011a), information retrieval (Al-Mubaid

& Nguyen, 2006; Budanitsky & Hirst, 2006b; Mohameth-François et al., 2012;

Pedersen et al., 2007), information integration (Petrakis et al., 2006; Rodríguez &

Egenhofer, 2003b) information extraction (IE) (Sánchez et al., 2011; Vicient et al.,

2013) and knowledge-based systems (Al-Mubaid & Nguyen, 2009).

Similarity is derived from the word similar (adjective), similarity (noun),

similitude (noun) and from the latin, similes, that is defined by like, resembling, or

similar (Harper, 2016). In mathematics, the word may be encountered in several

different contexts. In algebra, the terms that contain the same power of the variables

involved are said to be like, or have similar terms. Terms that are not similar are

called dissimilar. In geometry, two figures are said to be similar if they have the

same shape, though not necessarily the same size (Schwartzman, 1996).

The most popular method used to compare two concepts is via similarity. In

this method, similarity is used as a precondition to create interoperability between

agents or words by using different sources (Ehrig & Staab, 2004). Similarity using

mailto:[email protected]

2

semantics (semantic similarity) identifies similarity based on the likeliness of their

meaning or their related information (Yuhua et al., 2003; Martinez-Gil & Aldana-

Montes, 2013). Semantic similarity improves the understanding of textual resources

and increases the accuracy of knowledge-based applications (Jiang et al., 2015).

Most items dealing with semantic similarity have been developed using

taxonomies and ontologies (Batet et al., 2013; Budanitsky & Hirst, 2006; Couto et

al.,2007; Cross et al., 2013; Liu et al., 2012; Rodriguez & Egenhofer, 2003; Sánchez

& Batet, 2013; Sánchez et al.,2010; Sánchez et al., 2012). Ontology is of great

interest to the semantic similarity research community as it offers a formal

specification to a shared conceptualization. Several approaches to ontology-based

semantic similarity have been proposed. The approaches can be classified into a

structure-based approach, an information content-based approach, a feature-based

approach and a hybrid-approach. They are used either in single ontology or multiple

ontologies (Elavarasi et al., 2014; Saruladha et al., 2011a). Despite these approaches,

most semantic similarities are used to compute the similarity between concepts

within a single ontology and are rarely used in multiple ontologies (Rodríguez &

Egenhofer, 2003b, Petrakis et al., 2006, Batet et al., 2013). The use of a single

ontology does not ensure complete integration across a heterogeneous knowledge

system. With an increasing problem in integrating heterogeneous knowledge sources,

it is more dire that semantic similarity via multiple ontologies is studied. The

exploitation of multiple ontologies would also provide additional knowledge that can

improve similarity estimation and solve cases where terms are not represented in an

individual ontology. This is especially interesting within domains of knowledge such

as biomedicine where several big and detailed ontologies are available and offer

overlapping and complementary knowledge to similar topics (Al-Mubaid & Nguyen,

2009).

The following section in this chapter discusses the problem background in

semantic similarity. The research objectives and research questions for each

objective are explained and outlined followed by the scope and significance of this

research and the overview of the organization of this thesis.

3

1.2 Problem statement

Most semantic similarity methods are used to compute the semantic distance between

concepts within a single ontology, but is rarely used for multiple ontologies.

Semantic similarity in multiple ontologies is necessary for information integration

and retrieval. However, in order to relate to this, several factors need to be

considered. The first factor defines the scope of the matching problem (Sánchez et

al., 2012). Each ontology is built according to the expertise of the engineer’s point of

view which leads to the heterogeneity of the ontology. Ontological concepts rarely

constitute a perfect fit because of this heterogeneity and lack of consensus. In order

to integrate different knowledge on ontology, one has to consider that ontologies may

represent the same knowledge within the ambiguity of language. Most related works

rely on terminological matching of the concept label. However, this approach tends

to underestimate the real similarity between concepts due to differentiating text

labels. Moreover, since identifying the concept heavily relies on terminological

matching, sometimes concepts with differing labels that have the exact same

definitions and the potential to obtain equivalent concepts may be omitted because

the commonality concept may not have been properly evaluated. Figure 1.1 shows

the problem for this type of situation. For example, the concepts compared are

antibiotic in WordNet and anti-bacterial agent in MeSH. The identical terminology

for the paired concepts antibiotic and anti-bacterial agent is the concept of ‘Drug’.

However, this terminology underestimates the real similarity concept. This similarity

concept pair is closer to bactericide in WordNet and anti-bacterial agent in MeSH.

The ‘Drug’ concept is more of a general subsumer compared to both bactericide and

anti-bacterial agent. The second factor in dealing with ontology integration is to find

the similar equivalents of the concept that can act as the least common subsumer

(LCS). Previous research has used the taxonomic ancestry to identify the LCS (Batet

et al., 2012). However, this approach’s limitation is that it only uses the concept of

ancestry to find the LCS.

4

Figure 1.1: Example of terminological matching between subsumer of antibiotic

(WordNet) and antibacterial agent (MeSH) (source: Solé-Ribalta et al., 2014).

The third factor is related to the accuracy of similarity. The measurement of

similarity plays a crucial role in determining the similarities between concepts

(Sánchez et al., 2012). The similarities between concepts can create accurate

information. The ontology structure has been widely employed in similarity

measurement, especially on the semantic similarity single ontology, but not in

multiple ontologies. The semantic similarity in multiple ontologies inappropriately

uses the ontology structure in similarity measurements because the concept pair

consists of two different ontologies that have different structures whereas the

structure ontology cannot be compared with directly (Petrakis et al., 2006). Besides

that, the measurement structure-based ontology generally considers the shortest path

between concept pairs. Consequentially, it is unsuitable for wide and detailed

ontologies such as WordNet. As a result, several taxonomical paths are not taken into

consideration. Besides that, with the use of structure ontology, other features are

omitted as these features influence the semantic concept (e.g. common and non-

common concepts). ‘Feature-based’ is an approach that overcomes the limitation of

structure ontology. The method of this approach has more potential use in similarity

Entity

Physical entity

Causal agent

Agent

Drug

Medicine

Bactericide

e

Antibiotic

MeSH concept

Drug

Chemical actions and uses

Pharmacologic actions

Therapeutic uses

Anti-infective agents

Anti-bacterial agent

Terminological

matching

Semantic

similar

WordNet MeSH

5

measurements of multiple ontologies. This method considers measurements between

sets of features (Sánchez et al., 2012). However, some weaknesses were found in this

approach as the measurements are very limited in their applicability ontology

wherein this information is available. Another problem is that this approach depends

on the weighting parameter that balances the contribution of each feature (Rodríguez

& Egenhofer, 2003b). Only Petrakis et al., (2006) did not depend on the weighting

parameters. The maximum value is taken when the similarity synonym is more than

zero (similarity synonym > 0 =1). Due to this, the contribution resulting from other

features are omitted as sometimes, this feature has high potential in similarity

measurement. Besides that, by assuming a similarity value of more than zero to one,

this method will yield an unreliable result.

1.3 Research objectives

The goal of this research is to develop a feature-based semantic similarity in order to

identify concepts that are similar from multiple ontologies. This research covers:

(i) To develop the taxonomy of semantic similarity that contributes

understanding towards current approaches, issues and data related to the

topic.

(ii) To propose and evaluate a ontological feature algorithm in semantic

similarity for multiple ontologies in terms of feature-matching accuracy.

(iii) To develop and evaluate a feature-based mechanism to measure semantic

similarity of multiple ontologies to increase the accuracy of similarity.

1.4 Research questions

Some research questions were developed based on the research objectives:

The research question for research objective 1:

(i) What is the appropriate approach for semantic similarity with multiple

ontologies in order to understand the issues and data involved?

6

Research questions for research objective 2

(i) Does the feature-based measurement used is appropriate in measuring the

ontological features algorithm proposed?

(ii) Does the proposed ontological features algorithm increase the accuracy of

feature matching?

(iii) Is there any relationship between dependent variable (human scored) and

independent variables (similarity values for each method)?

Research questions for research objective 3

(i) How do the proposed parameters support in improving the accuracy of

similarity for multiple ontologies?

(ii) Does it a significant relationship between the Hyb-TvX methods with human

scored?

1.5 Scope and research significance

This research, focused on the semantic similarity between multiple ontologies. The

datasets used were WordNet and MeSH. The WordNet dataset was downloaded from

https://wordnet.princeton.edu whereas the MeSH dataset was downloaded from

http://www.nlm.nih.gov/mesh/MBrowser.html. This research will try to improve the

ontological features algorithm. The ontological features algorithm consists of

matching matrix semantic relationships where this component has a matching matrix

of a subsumer. The matching matrix semantic relationships uses the semantic

relationship matrix which is divided into four partitions; hypernym, hyponym, sister

term and meronym/holonym. After the semantic relationship matrix is defined, the

subsumer of each relationship in the phase matching matrix of subsumer is identified.

Besides that, the ontological features algorithm also includes the semantic

overlapping between subsumers where a matching matrix of subsumer computes to

subsumers according to the number of hyponyms that they share. The last phase is

selecting the most suitable and least common subsumer (LCS).

This research proposed the Hyb-TvX which is a hybrid of the X-similarity

and Tversky methods. The proposed Hyb-TvX method proved its capability in

dataets for different ontologies in the biomedical domain. The Hyb-TvX consists of

http://www.nlm.nih.gov/mesh/MBrowser.html

7

two components: similarity measurement level 1 (TvX-1) and similarity

measurement level 2 (TvX-2). The TvX-1, has two phases: (i) calculation of concept

to find the similarity between concepts and (ii) calculation of synonyms to find the

similarity synonym of each concept. The TvX-2 has two phases: (i) calculation of

features to find a similarity concept through features and (ii) normalization of all

calculations to find the similarity value for that concept. The evaluation measurement

of Hyb-TvX encompasses the similarity of each concept pair as compared with the

previous similarity method, the measurement evaluation (correlation) and the p-

value. The focus of this research is to develop a better semantic similarity

measurement method in order to obtain optimum accuracy.

This research could contribute to the "body of knowledge" in a feature-based

semantic similarity to support multiple ontologies and the biomedical domain. This

research also contributes to the improvement of statistical publications and citations

for a number of articles in the areas of research. Through the web of science (12

January 2017), Figure 1.2 shows that the research semantic similarity for multiple

ontologies and the feature-based semantic similarity is growing in importance and

subsequently gaining more attention from other researchers around the world.

However, research in the field of semantic similarity feature-based approach using

multiple ontologies still needs to be explored by the researcher.

Furthermore, the results of this research could also act as a guide for

expanded research in the field of semantic similarity with multiple ontologies using a

feature-based approach that has, only previously received little attention by

researchers. This research is also able to resolve the issues in the terminological

matching method with the developed ontological features algorithm. An analysis of

the similarity value and correlation indirectly provided a clear picture that the

proposed ontological features algorithm can improve the similarity and correlation

values that, in turn, can be used as empirical evidence. The issue of a similarity

measurement that depends on the weighting parameter that balances the contribution

of each feature (Rodríguez & Egenhofer, 2003b) and other features that are omitted

can be addressed by developing the proposed method (Hyb-TvX). Alongside this, an

analysis of the similarity, correlation and p-value will also indirectly confirm that the

proposed method (Hyb-TvX) can increase the similarity, correlation and would be

capable of showing a significant correlation. Overall, this analysis aims to obtain

optimum accuracy of similarity between the concepts.

8

Additionally, the results of semantic similarity using the biomedical domain

is important in assisting with the integration of heterogeneous clinical data such as

clinical records with different formats. This research can improve the interoperability

between medical sources, which are commonly dispersed and standalone. Therefore,

it can assist in searching within the biomedical domain and ensuring that such

searches are accurate.

The number of articles published per year The number of citations per year

Semantic similarity with multiple ontologies

Semantic similarity feature-based

Semantic similarity with multiple ontologies feature-based approach

Figure 1.2: Comparison of the number of articles published every year and the

number of citations per year (http://apps.webofknowledge.com)

9

1.6 Organization of the thesis

This thesis is organized in six chapters. Chapter 1 describes the introduction,

problem statements, research objectives, research questions, scope and significance

of this research. The rest of the thesis is organized into the following chapters.

Chapter 2 reviews the main subjects used in the research which includes similarity,

the approach of measurement analysis for semantic similarity and a feature-based

semantic similarity approach for multiple ontologies. Chapter 3 describes the design

of the semantic similarity method adopted to achieve the objectives of the research.

This includes the research framework, data sources, instrumentation and result

analysis. Chapter 4 highlights the development of the ontological features algorithm

that aims to find the correspondence between compared concepts. This algorithm

will describe the ontological features used in the similarity measurement method in

the next chapter. Chapter 5 further elaborates the development of the feature-based

Hyb-TvX method that utilizes a hybrid X-similarity and Tversky method. This

method also includes ontological features and the use of a semantic relationship of

concepts such as: hypernym, hyponym, sister term and meronym/holonym. This

chapter also elaborates on the evaluation of Hyb-TvX with other methods in feature-

based approaches. Finally, chapter 6 presents the general conclusions of the research,

the contribution and proposed topics for future research.

2CHAPTER 2

LITERATURE REVIEW

2.1 Introduction

This chapter strives to give a better understanding of similarity while reviewing

several works on semantic similarity. This chapter also explores the ontological

features matching method for enabling semantic similarity from multiple ontologies.

Furthermore, approaches to the measurement analysis of semantic similarity for

multiple ontologies are discussed. The current research trends and directions are

outlined before presenting the summary of this chapter. The content structure is

represented in Figure 2.1.

.

Figure 2.1: Content structure

Section 2.10

Section 2.11

Section 2.9

Discussion of

Similarity

Summary

Biomedical

domain

Single

Ontology

Section 2.6

Section 2.7

Section 2.8

Classification

of Semantic

Similarity Multiple

Ontologies

Section 2.5.1

Ontology

General

purpose

ontology

Specific

ontology

Section 2.5.2 Section 2.5.3

Section 2.4 Section 2.5 Section 2.2

Similarity

Section 2.3

Words

Similarity

Semantic

Similarity

Knowledge-

based

11

2.2 Similarity

The first issue addresses the meaning of similarity. Generally, similarity is a quality

or condition of being similar. However, many different definitions of similarity are

possible, each being appropriate for specific and particular situations. This statement

is also supported by Lin (1998a), who opines that the definition of similarity is

normally according to the application and representation of knowledge. Similarity is

also defined as a basis in making predictions, because similar things usually behave

similarly (Quine, 1969). According to Lin (1998a), three formal definitions of the

concept of similarity exist:

(i) Definition by Concept 1: Similarity between x and y relates to their

commonality. The more commonality they share; they are more similar.

(ii) Definition by Concept 2: Similarity between x and y relates to the differences

between them. The more differences they have, they are less similar.

(iii) Definition by Concept 3: The maximum similarity between x and y is reached

when x and y are identical, no matter how much commonality they share.

Definition 2.1 (Similarity): A similarity σ : oo →R is a function from a pair of

entities to real number (R) expressing the similarity between two objects (o and o).

Table 2.1 shows the symbols that describe the meaning of similarity according to

definition 2.1:

0,,, yxoyx sspositivene

zyxxozyox ,,,,, imalitymax

xyyxoyx ,,,, symmetry

12

Table 2.1: Symbol and meaning in relation to similarity

Symbol Meaning

for all,

for any,

for each,

for every

is an element of,

is not an element of

selection of

is less than or equal to,

is greater than or equal to

2.3 Words similarity

In many fields of research on similarity, the use of the word has solved a lot of

problems and has also given benefits in associated fields such as text categorization

(Ko et al., 2004), text summarization (Erkan & Radev, 2004; Lin & Hovy, 2003),

word sense disambiguation (Lesk, 1986; Schütze, 1998), automatic evaluation of

machine translation (Liu & Zong, 2004; Papineni et al., 2002), evaluation of text

coherence (Lapata & Barzilay, 2005; Wegrzyn-Wolska & Szczepaniak, 2005) and

the classification of formatted documents (Wegrzyn-Wolska & Szczepaniak, 2005).

Cohen (2000) stated that word similarity is vital in the retrieval of images from the

web. This can improve the retrieval of images by utilising informative words.

Word similarity is used as the primary stage to assess the similarities between

sentence, paragraph and document. This statement is supported by Lin (1998b)

where similarity between two documents can be calculated by comparing the sets of

concept in the documents or by comparing their stylistic parameter values, such as

average word length, average sentence length, and average number of verbs per

sentence. Several studies have assumed that words which are close in meaning will

occur in similar pieces of text and context (Gomaa & Fahmy, 2013; Kolb, 2009;

Landauer & Dumais, 1997; Lin, 1998b).

Word similarity in the context databases can be used in schema matching to

solve semantic heterogeneity. The main problem regarding similarity in the context

of a database is the data sharing system; whether it is a federated database, a data

integration system, a message passing system, a web service, or a peer-to-peer data

management system (Madhavan et al., 2005). Word similarity also involves a joint

operator as it joins two relations if their attributes are textually similar to each other.

13

Besides that, it also has a variety of application domains including integration and

querying of data from heterogeneous resources, cleansing of data and mining of data

(Cohen, 2000; Islam & Inkpen, 2008; Schallehn et al., 2004). From the current

studies, there are two approaches for word similarity. The first approach is similarity

from a lexical perspective and the second approach is from the semantic approach.

The lexical approach is referred to as words that are similar if they have a

similar character sequence. In this research, the lexical approach is introduced

through the different string-based method. The string-based method works on string

sequences and character structure (Bernstein & Rahm, 2001). This method typically

finds the concepts of ‘Book’ and ‘Textbook’ to be similar, but not the concepts of

‘Book’ and ‘Volume’. The string-based method is often used to match names and

their description. This method assumes that the more similar the string, the more

likely they are in representing the same concept (Bernstein & Rahm, 2001; Gomaa &

Fahmy, 2013). There are many ways to compare the string depending on the way the

string is viewed, for example as an exact sequence of letters, a set of letters, and a set

of words (Euzenat & Shvaiko, 2007). There are two approaches of the string-based

method identified in this research: character-based and term-based.

(i) The character-based approach considers distance as the difference between

the characters. This is useful in the case of typographical errors. Among the

works that have used this type include the Longest Common Substring

(Allison & Dix, 1986), Damerau (1964), Jaro (1995), Winkler (1990),

Needleman & Wunsch (1970), Smith & Waterman (1981) and N-gram

(Kondrak, 2005).

(ii) The term-based approach is a similarity measure which incorporates the

linguistic and semantic structures using syntactic dependencies. This type

comes from information retrieval and considers a string as a multi set of

words. These approaches usually work well on long texts (comprising of

many words). This approach can also adapt to ontology concepts such as

aggregating different sources of string, example identifiers, labels, comments

and documentation. Besides that, this approach can also split the string into

independent tokens. For example, ‘renal failure’ becomes ‘renal’ and

‘failure’. Examples of concepts that used this approach are Jaccard (1901)

and Dice (1945).

14

The second approach is semantic. Semantic refers to words that can be

similar if they have the same thing, used in the same way and in the same context

(Gomaa & Fahmy, 2013). There are two types identified in this semantic: similarity

and relatedness.

(i) Similarity or better known as semantic similarity is a comparison among

entities/terms/concepts. This semantic similarity allows information retrieval

and information integration to handle concepts that are semantically similar.

Example of works that have used semantics are Resnik (1995), Jiang &

Conrath (1997), Palmer & Wu (1994), Leacock & Chodorow (1998),

Rodríguez & Egenhofer (2003a), Saruladha et al., (2011b) and Sánchez et al.,

(2012).

(ii) Relatedness, or otherwise known as semantic relatedness is a more general

notion of relatedness, not specifically tied to the shape or form of the concept

and they are not limited to considering is-a relations (Gomaa & Fahmy,

2013). Among works that have used this type are the hso measure (Hirst &

St-Onge, 1998), lesk (Banerjee & Pedersen, 2002) and vector pair

(Patwardhan & Pedersen, 2006).

Based on the number of reviews, the lexical approach is an easy method

because it measures string sequences and character composition. However, a

limitation on this approach exists, where it is considered as a simple traditional

method in determining the similarity of concepts (Islam & Inkpen, 2008). This

approach cannot identify the semantic similarity of concept. For instance, with the

similarity between the concepts of bactericide and anti-bacterial agent, the current

word similarity is unsuccessful in identifying any kind of connection between these

words. The limitation of the lexical approach can be overcome by using the semantic

approach. The semantic approach does not rely on the string sequence, but also

measures similarity based on the likeness between words. This approach can identify

the same meaning in different words. However, this approach needs an evaluation of

the semantic evidence observed in knowledge sources such as ontologies or domain

corpora. Most knowledge sources are presented in unprocessed and heterogeneous

textual formats (Batet et al., 2011). Table 2.2 displays the conclusion of words

similarity approach as previously described.

15

Table 2.2: Words similarity approach

Approach Techniques Example Research Advantages Disadvantages

Lexical

Character-based

Allison & Dix (1986); Damerau (1964);

Jaro (1995); Winkler (1990); Needleman &

Wunsch (1970); Smith & Waterman (1981)

and Kondrak (2005).

Easy method because this approach

measures on the string sequences

and character composition.

It is useful if use very similar string

to denote the same concepts.

These approaches are most often

used in order to detect very similar

string used.

Different concepts with different

structure characters are used, this

will yield a low similarity.

Concept pair with low similarity

in turn yields many false positive.

Term-based Jaccard (1901); Dice (1945)

Semantic

Similarity

Resnik (1995); Jiang & Conrath (1997);

Palmer & Wu (1994); Leacock & Chodorow

(1998); Rodríguez & Egenhofer (2003a);

Saruladha et al., (2011b) and Sánchez et al.,

(2012)

This approach does not depend on

the sequence of string.

Semantic similarity computes the

likeness between words, understood

as the degree of taxonomical

proximity.

Need evaluation of the semantic

evidence observed in knowledge

source.

Knowledge source presented in

unprocessed and heterogeneous

textual formats Relatedness

Hirst & St-Onge (1998); Banerjee &

Pedersen (2002) and Patwardhan &

Pedersen (2006).

16

2.4 Semantic similarity

Basically, semantic similarity is the quality or condition of being similar. The

differences being purely dependant on certain situations. Doan et al., (2004)

mentioned that semantic similarity can be defined based on the joint probability

distribution of the concepts involved. According to Elavarasi et al., (2014) semantic

similarity is defined as the closeness of two concepts based on the likeliness of their

meaning, which refers to the similarity between two concepts in a taxonomy or

ontology. Besides that, according to Jiang et al., (2015) semantic similarity relates to

computing the similarity between concepts, words, terms or short text expressions,

where the concepts have the same meaning or have relatively matching information

despite not being lexically similar (Martinez-Gil & Aldana-Montes, 2013; Yuhua et

al., 2003).

According to Schwering (2008), semantic similarity is central for the proper

function of semantically enabled processing of geospatial data. It is used to measure

the degree of potential semantic interoperability between data or geographic

information systems (GIS). This semantic similarity is used to deal with vague data

queries, vague concepts or natural language. This is also supported by Zhang et al.,

(2015) whom indicated that semantic similarity is the degree of semantic equivalence

between two linguistic items, where the items can be concepts, sentences or

documents. However, semantic similarity between sentences or documents can also

be known as semantic textual similarity (STS).

Semantic similarity has been used for years in psychology and cognitive

science where different models have been proposed (Pirró & Euzenat, 2010). Besides

that, semantic similarity has also been applied in searching for similarities between

images and visual media (Deselaers & Ferrari, 2011). However, in recent years,

semantic similarity is widely used in obtaining similarities between concepts or

words where it assists in information extraction tasks (Sánchez & Isern, 2011) such

as semantic annotation (Sánchez et al., 2011), and ontology learning (Iannone et al.,

2007).

According to Schwering (2008), semantic similarity is also widely used in

information retrieval tasks (Al-Mubaid & Nguyen, 2009; Budanitsky & Hirst, 2006b;

17

Saruladha et al., 2011a) to improve the performance of current search engines

(Hliaoutakis et al., 2006), information integration (Saruladha et al., 2011a), ontology

matching (Pirrò et al., 2009, Saruladha et al., 2011a), semantic query routing, and

bioinformatics to assess the similarity between proteins (Wang et al., 2007).

Additionally, semantic similarity can also play an important role in both predicting

and validating gene product interactions and interaction networks (Pesquita et al.,

2009). Table 2.3 shows the field and application that utilises the semantic similarity

approach.

Table 2.3: Fields and applications that use the semantic similarity approach

Approach Field Application Reference

Semantic Similarity

Information extraction

Semantic annotation Sánchez & Isern,

(2011); Sánchez et

al., (2011); Iannone

et al., (2007) Ontology learning

Information retrieval

Performance search

engine Al-Mubaid &

Nguyen (2009);

Budanitsky & Hirst

(2006b); Saruladha et

al., (2011a);

Hliaoutakis et al.,

(2006); Pirrò et al.,

(2009); Wang et al.,

(2007)

Information integration

Ontology matching

Semantic query rating

Bioinformatic

Semantic similarity is generally based on certain background information

(Maind et al., 2012). Two background information are identified in semantic

similarity: structured information and unstructured information (Abdelrahman &

Kayed, 2015).

(i) Structured information is often in a hierarchical form that is known as

knowledge-based such as the WordNet (Miller, 1995), MeSH (Hliaoutakis et

al., 2006), the Systemized Nomenclature of Medicine Clinical Term

(Snomed-CT) (Garla & Brandt, 2012), Wikipedia (Gabrilovich &

Markovitch, 2007) and Gene Ontology (GO) (Ashburner et al., 2000).

Similarity measurement from knowledge-based determines the degree of

similarity between texts using information derived from semantic networks.

18

Examples of similarity using knowledge-based are found in Resnik (1995),

Jiang & Conrath (1997), Palmer & Wu (1994) and Leacock & Chodorow

(1998).

(ii) Unstructured information refers to corpus-based, which are a collection of

texts. Some examples that use a corpus-based includes the Brown Corpus and

Wall Street Journals (Zhang et al., 2015). Similarity measurement from

corpus-based defines the similarity between texts as dependant on the

information gained from the large corpora. A corpus is a large collection of

written or spoken texts that is used for language research. There are several

studies using corpus-based; for example, Gabrilovich & Markovitch (2007),

Landauer & Dumais (1997) and Lund et al., (1995).

Based on a number of reviews, the corpus-based unstructured information

represents the semantic of words by distribution in large multilingual corpora. These

unstructured information rely on the assumption that related words exist in the same

document (Aggarwal, 2012). However, similarity that uses corpus-based performs

well in document similarity, but needs further improvement for short text or phrases.

Knowledge-based is one potential issue in semantic similarity, especially for

applications dealing with textual data (Batet et al., 2014; Pirró & Euzenat, 2010)

where semantic similarity measurement between words provides a valuable tool to

the understanding of textual resources (Sánchez et al., 2012). According to Batet &

Sánchez (2014), semantic similarity measurement that uses knowledge-based is able

to capture the semantics inherent to the knowledge modelled in ontology. Two types

of knowledge are exploited; (i) explicit knowledge such as the structure of a

taxonomy and (ii) implicit knowledge such as information distribution (Sánchez et

al., 2010). However, knowledge-based semantic similarity needs a proper

understanding of the semantics concept. Encouraging an improved use and

integration of heterogeneous sources as well as higher information retrieval accuracy

(Saleena & Srivatsa, 2014) are some of the ways that will help improve

understanding. Table 2.4 depicts the advantages and disadvantages based on the

background information of the semantic similarity approach.

19

Table 2.4: Advantages and disadvantages based on the background information in

the semantic similarity approach

Approach Background

Information Advantage Disadvantage

Semantic

similarity

Structure Information

(knowledge-based)

Determines the degree of

similarity between texts

using information derived

from semantic networks.

Capable to capture the

semantics inherent to the

knowledge model led in

ontology.

Need proper understanding

of concept semantics

Unstructured

Information (corpus-

based)

Similarity used corpus-

based perform well in

document similarity.

Rely on the assumption that

related words exist in the

same document.

Need to improve for short

text or phrases.

2.5 Knowledge-based

Knowledge-based similarity determines the degree of similarity between concepts

using information derived from semantic networks. WordNet (Fellbaum, 1998) is the

most common semantic network in the area of measuring the knowledge-based

similarity between concepts. Knowledge-based uses ontology when calculating

similarity (Batet, et al., 2013). The reason ontologies are so popular is due, in large

part, to what they promise: a shared and a common understanding of some domains

that can be communicated across people and computers (Studer et al., 1998).

Ontology is a type of knowledge-based. Ontology describes concepts through

definitions that are sufficiently detailed to capture the semantics of a domain.

Ontologies are widely used to enrich the semantics of the web (Alasoud, 2009).

2.5.1 Ontology-based

“An ontology is defined as a formal, explicit specification of a shared

conceptualisation” which means that ontology is defined as a formal representation

of concepts within a domain and the relationship between those concepts (Studer et

al., 1998). Ontology is an effective way to share knowledge within controlled and

structured vocabulary (Spasic et al., 2005). Many ontologies have been developed

20

for various purposes and domains (Al-Mubaid & Nguyen, 2009; Hliaoutakis, 2005,

Miller, 1995). Furthermore, in reference to Noy & McGuinness (2001) ontology is

built for some reasons such as sharing a common understanding of the structure of

information among people or software agents, enabling the reuse of domain

knowledge, making explicit domain assumptions, separating the domain knowledge

from the operational knowledge and analysing domain knowledge. Besides that,

ontology is also crucial in enabling interoperability across heterogeneous systems

and semantic web applications (Choi et al., 2006).

Ontology contains concepts, the definitions of these concepts, and rich

relationships among these concepts. Consider the computer ontology example shown

in Figure 2.2.

Figure 2.2: The computer ontology (Alasoud, 2009)

Three basic components of ontology are:

(i) Concept or classes

These are concepts of the domain or task, usually organised in taxonomies. In

our ontology example, Computer, PC, Laptop, Hard Disk, etc. are examples

of the concept.

(ii) Roles or properties

These are the types of interaction between instances of concepts in the

domain. For example, has-HD (has Hard Disk), has-monitor, and has-maker

are roles which are shown in Figure 2.2.

(iii) Individual or Instances

Individuals or instances represent specific elements.

is-a Computer

Laptop

Hard Disk Monitor

PC

Used-PC Maker

is-a

has-HD has-monitor has-maker is-a

21

Ontology is a type of knowledge-based that describes concepts through

definitions that are sufficiently detailed to capture the semantics of a domain. A few

ontologies such as the WordNet (Miller, 1995) have been used for semantic

similarity. The WordNet is a lexical database for general English covering most

generic English concepts and supports various purposes. Besides that, other

ontologies are also used for the same purpose as the Unified Medical Language

System (UMLS), that includes many biomedical ontologies and terminologies (e.g.,

MeSH, Snomed-CT) (Saruladha et al., 2011b), and the International Classification

Disease (ICD) family (Al-Mubaid & Nguyen, 2009). These ontologies are

specifically created for the biomedical domain that is different from WordNet.

Ontology-based semantic similarity is used in two situations. The semantic

similarity in a single ontology and when multiple ontologies are involved.

(i) Single ontology means similarities are compared from the same ontology, an

example is illustrated in Figure 2.3.

(ii) Multiple ontologies mean that the similarity concepts are compared from

different ontologies, example is illustrated in Figure 2.4.

22

Figure 2.3: Semantic similarity in single ontology for male and female WordNet ontology (Yatskevich & Giunchiglia, 2007)

WordNet

Ornamental

Plant

Living Thing Whole, Unit

Physical Object

Physical Entity

Entity

Artifact Organism

Instrumentation Sphere

Furnishing

Bed Table

Furniture

Celestial Globe

Globe

Person

Male Female

Legend: is-a relationship

other connections

semantic similarity

connection

23

Figure 2.4: Semantic similarity in multiple ontologies for intracranial hemorrhage (in Snomed-CT) and brain neoplasms (in MeSH) (Batet et al.,

2014)

Snomed-CT

Snomed-CT

Clinical finding

Bleeding Finding by site Disease

Observation of body region Disease by body site

Head and neck finding

Head finding

Disorder of body cavity (disorder)

Disease of head

Intracranial haemorrhage

MeSH

Thing

Disease

Nervous system diseases

Central nervous system diseases

Brain diseases

Neoplasms

Neoplasms by site

Nervous system neoplasms

Central nervous system neoplasms

Brain neoplasms

Legend:

is-a relationship

semantic similarity

connections

24

2.5.2 General purpose ontologies used with semantic similarity approaches

There are several examples of general purpose ontologies available including:

WordNet, SENSUS, and the Cyc knowledge base. The following section describes

the general purpose ontologies as follows:

2.5.2.1 WordNet

WordNet is the lexical knowledge of a native speaker of English. The latest version

of WordNet is v3.1 which was released in June 2011. WordNet has 117,659 synsets

and 206,941 general concepts of different domains (Slimani, 2013). These databases

are semantically structured in ontological ways. It also contains nouns, verbs,

adjectives and adverbs that are linked to synonym sets (synset), where each synset

consists of a list of synonym word forms and semantic pointers that describe the

relationships between the current synset and other synsets (Hliaoutakis et al., 2006).

Different types of relationships can be derived between the synsets or concepts

(related to other synsets higher or lower in the hierarchy).

The hyponym/hypernym relationship (i.e., is-a relationship), and the

meronym/holonym relationship (i.e., part-of relationship) are the most recognized

relationships in WordNet. WordNet also introduces a larger amount of abstract

concepts at the top of the taxonomic tree (Solé-Ribalta et al., 2014). This is due to it

being a general lexical database that does not merely focus on a singular domain.

Figure 2.5 denotes the snapshot of WordNet web pages. The WordNet typically

displays information such as synonym (S), direct hyponym (children of concept),

direct hypernym (direct parents), full hyponym (all children), inherited hypernym (all

parents), and sister term (shared direct parents). The WordNet contains a description

of the concept in the form of tree structure as displayed in Figure 2.6.

REFERENCES

Abdelrahman, A. M. B., & Kayed, A. (2015). A Survey on Semantic Similarity

Measures between Concepts in Health Domain. American Journal of

Computational Mathematics, 5(June), 204–214. https://doi.org/10.4236/

ajcm.2015.52017.

Aggarwal, N. (2012). Cross lingual semantic search by improving semantic

similarity and relatedness measures. Lecture Notes in Computer Science

(Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes

in Bioinformatics), 7650 LNCS, 375–382. https://doi.org/10.1007/978-3-642-

35173-0-26.

Al-Mubaid, H., & Nguyen, H. A (2009). Measuring semantic similarity between

biomedical concepts within multiple ontologies. IEEE Transactions on Systems,

Man and Cybernetics Part C: Applications and Reviews, 39(4), 389–398.

Retrieved from http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5061528

Al-Mubaid, H., & Nguyen, H. A. (2006). A cluster-based approach for semantic

similarity in the biomedical domain. In Annual International Conference of the

IEEE Engineering in Medicine and Biology - Proceedings (pp. 2713–2717).

https://doi.org/10.1109/IEMBS.2006.259235.

Alasoud, A. K. (2009). A Multi-Matching Technique for Combining Similarity

Measures in Ontology Integration. Concordia University School of Graduate

Studies. Concordia University. Thesis-PhD

Allison, L., & Dix, T. I. (1986). A bit-string longest-common-subsequence

algorithm. Information Processing Letters. https://doi.org/10.1016/0020-

0190(86)90091-8.

Ashburner, M., Ball, C. A., Blake, J. A., Botstein, D., Butler, H., Cherry, J. M., &

Al., A. P. D. (2000). Gene Ontology: tool for the unification of biology. Nature

Genetics, 25(1), 25–29. https://doi.org/10.1038/75556.

Banerjee, S., & Pedersen, T. (2002). An Adapted Lesk Algorithm for Word Sense

146

Disambiguation Using WordNet. International Conference on Intelligent Text

Processing and Computational Linguistics, 2276, 136–145.

https://doi.org/10.1007/3-540-45715-1_11.

Batet, M., Harispe, S., Ranwez, S., Sánchez, D., & Ranwez, V. (2014). An

information theoretic approach to improve semantic similarity assessments

across multiple ontologies. Information Sciences, 283, 197–210.

https://doi.org/10.1016/j.ins.2014.06.039.

Batet, M., & Sánchez, D. (2014). A Semantic Approach for Ontology Evaluation.

2014 IEEE 26th International Conference on Tools with Artificial Intelligence,

138–145. https://doi.org/10.1109/ICTAI.2014.30.

Batet, M., Sánchez, D., & Valls, A. (2011). An ontology-based measure to compute

semantic similarity in biomedicine. Journal of Biomedical Informatics, 44(1),

118–125. https://doi.org/10.1016/j.jbi.2010.09.002.

Batet, M., Sánchez, D., Valls, A., & Gibert, K. (2010). Exploiting taxonomical

knowledge to compute semantic similarity: An evaluation in the biomedical

domain. Lecture Notes in Computer Science (Including Subseries Lecture Notes

in Artificial Intelligence and Lecture Notes in Bioinformatics), 6096 LNAI, 274–

283. https://doi.org/10.1007/978-3-642-13022-9_28.

Batet, M., Sánchez, D., Valls, A., & Gibert, K. (2013). Semantic similarity

estimation from multiple ontologies. Applied Intelligence, 38(1), 29–44.

https://doi.org/10.1007/s10489-012-0355-y.

Bernstein, P. A., & Rahm, E. (2001). Generic Schema Matching , Ten Years Later,

695–701.

Blank, A. (2003). Words and concepts in time: Towards diachronic cognitive

onomasiology. Trends In Linguistics Studies And Monographs, 6–25. Retrieved

from http://www.metaphorik.de/01/blank.htm.

Bodenreider, O. (2004). The Unified Medical Language System (UMLS): integrating

biomedical terminology. Nucleic Acids Research, 32(suppl 1), D267–D270.

https://doi.org/10.1093/nar/gkh061.

Budanitsky, A., & Hirst, G. (2006a). Evaluating WordNet-based Measures of Lexical

Semantic Relatedness. Computational Linguistics. https://doi.org/10.1162/

.2006.32.1.13.

Budanitsky, A., & Hirst, G. (2006b). Evaluating WordNet-based Measures of

Semantic Distance. Computational Linguistics, 32, 13–47.

147

https://doi.org/10.1162/coli.2006.32.1.13.

Bulskov, H., Knappe, R., & Andreasen, T. (2002). On measuring similarity for

conceptual querying. Flexible Query Answering Systems, 100–111. Retrieved

from http://link.springer.com/chapter/10.1007/3-540-36109-X_8.

Choi, N., Song, I.-Y., & Han, H. (2006). A Survey on Ontology Mapping, College of

Information Science and Technology Drexel University, Philadelphia, PA

19014 35(3): 34–41.

Cohen, W. (2000). Data integration using similarity joins and a word-based

information representation language. ACM Transactions on Information

Systems. https://doi.org/10.1145/352595.352598.

Couto, F. M., Silva, M. J., & Coutinho, P. M. (2007). Measuring semantic similarity

between Gene Ontology terms. In Data and Knowledge Engineering, 61, 137–

152. https://doi.org/10.1016/j.datak.2006.05.003

Cross, V., Yu, X., & Hu, X. (2013). Unifying ontological similarity measures: A

theoretical and empirical investigation. In International Journal of Approximate

Reasoning (Vol. 54, pp. 861–875). https://doi.org/10.1016/j.ijar.2013.03.003

Damerau, F. J. (1964). A technique for computer detection and correction of spelling

errors. Communications of the ACM. https://doi.org/10.1145/363958.363994.

Deselaers, T., & Ferrari, V. (2011). Visual and semantic similarity in ImageNet. In

Proceedings of the IEEE Computer Society Conference on Computer Vision and

Pattern Recognition (pp. 1777–1784). https://doi.org/10.1109/CVPR.2

011.5995474.

Dice, L. R. (1945). Measures of the Amount of Ecologic Association Between

Species. Ecology, 26, 297–302. https://doi.org/10.2307/1932409.

Ding, L., Finin, T., Joshi, A., Pan, R., Cost, R. S., Peng, Y., Reddivari, P., Doshi, V.,

Sachs, J.(2004). Swoogle : A Semantic Web Search and Metadata Engine. In

Proceedings of the thirteenth ACM international conference on Information and

knowledge management (pp. 652–659).

https://doi.org/10.1145/1031171.1031289.

Doan, A., Madhavan, J., Domingos, P., & Halevy, A. (2004). Ontology matching: A

machine learning approach. In Handbook on Ontologies, 1–20. Retrieved from

http://link.springer.com/chapter/10.1007/978-3-540-24750-0_19.

Ehrig, M., & Staab, S. (2004). QOM–quick ontology mapping. The Semantic Web–

ISWC 2004, 1–28. Retrieved from http://link.springer.com/chapter/10.1007/978-

148

3-540-30475-3_47.

Elavarasi, S., Akilandeswari, J., & Menaga, K. (2014). A Survey on Semantic

Similarity Measure. Ijrat.org, 2(3), 389–398. Retrieved from

http://www.ijrat.org/downloads/march-2014/paper id-232014114.pdf.

Erkan, G., & Radev, D. R. (2004). LexRank: Graph-based lexical centrality as

salience in text summarization. Journal of Artificial Intelligence Research, 22,

457–479. https://doi.org/10.1613/jair.1523.

Euzenat, J., & Shvaiko, P. (2007). Ontology Mapping. Springer Berlin Heidelberg.

Fellbaum, C. (1998). WordNet: An Electronic Lexical Database. British Journal Of

Hospital Medicine London England 2005 (Vol. 71).

https://doi.org/10.1139/h11-025.

Gabrilovich, E., & Markovitch, S. (2007). Computing Semantic Relatedness using

Wikipedia-based Explicit Semantic Analysis. Proceedings of the 20th

International Joint Conference on Artificial Intelligence (pp. 1–6).

https://doi.org/10.1145/2063576.2063865.

Garla, V. N., & Brandt, C. (2012). Semantic similarity in the biomedical domain: an

evaluation across knowledge sources. BMC Bioinformatics, 13, 261.

https://doi.org/10.1186/1471-2105-13-261.

Gomaa, W., & Fahmy, A. (2013). A survey of text similarity approaches.

International Journal of Computer Applications, 68(13), 13–18.

Grossman, J., Grossman, M., & Katz, R. (1980). The first system of weighted

differential and integral calculus, ISBN 0-9771170-1-4

Harper, D. (2016). Online Etymology Dictionary. Access at June 25, 2016, from

http:// http://www.etymonline.com/index.php?l=s&p=42

Hirst, G., & St-Onge, D. (1998). Lexical chains as representations of context for the

detection and correction of malapropisms. WordNet - An Electronic Lexical

Database, 305–332. https://doi.org/citeulike-article-id:4893262.

Hliaoutakis, A. (2005). Semantic Similarity Measures in MeSH Ontology and their

application to Information Retrieval on Medline. Technical University of Crete,

Greek, Thesis PhD.

Hliaoutakis, A, Varelas, G., Voutsakis, E., Petrakis, E. G. M., & Milios, E. (2006).

Information Retrieval by Semantic Similarity. International Journal on

Semantic Web and Information Systems, 2, 55–73.

https://doi.org/10.4018/jswis.2006070104.

149

Iannone, L., Palmisano, I., & Fanizzi, N. (2007). An algorithm based on

counterfactuals for concept learning in the Semantic Web. Applied Intelligence,

26, 139–159. https://doi.org/10.1007/s10489-006-0011-5.

Islam, A., & Inkpen, D. (2008). Semantic text similarity using corpus-based word

similarity and string similarity. ACM Transactions on Knowledge Discovery

from Data. https://doi.org/10.1145/1376815.1376819.

Jaccard, P. (1901). Étude comparative de la distribution florale dans une portion des

Alpes et des Jura. Bulletin de La Société Vaudoise Des Sciences Naturelles, 37,

547–579. https://doi.org/citeulike-article-id:4118068.

Jaro, M. A. (1995). Probabilistic linkage of large public health data files. Statistics in

Medicine, 14, 491–498. https://doi.org/10.1002/sim.4780140510.

Jean-Mary, Y. R., Shironoshita, E. P., & Kabuka, M. R. (2009). Ontology Matching

with Semantic Verification. Journal of Web Semantics, 7(3), 235–251.

https://doi.org/10.1016/j.websem.2009.04.001.

Jiang, J., & Conrath, D. (1997). Semantic similarity based on corpus statistics and

lexical taxonomy. arXiv Preprint Cmp-lg/9709008. Retrieved from

http://arxiv.org/abs/cmp-lg/9709008.

Jiang, R.., Gan, M., & Doy, X., (2013). From ontology to semantic similarity:

calculation of ontology-based semantic similarity. The Scientific World Journal,

2013, 793091. https://doi.org/10.1155/2013/793091.

Jiang, Y., Zhang, X., Tang, Y., & Nie, R. (2015). Feature-based approaches to

semantic similarity assessment of concepts using Wikipedia. Information

Processing and Management, 51(3), 215–234.

https://doi.org/10.1016/j.ipm.2015.01.001.

Kasim, S. (2011). Fuzzy C-Means Clustering By Incorporating Biological

Knowledge And Multi-Stage Filtering To Improve Gene Function Prediction.

University Teknologi Malaysia.Thesis-PhD

Kleinsorge, R., Tilley, C., & Willis, J. (2002). Unified Medical Language System

(UMLS). Encyclopedia of Library and Information Science, 369–378.

https://doi.org/10.1002/9781118479612.ch16.

Ko, Y., Park, J., & Seo, J. (2004). Improving text categorization using the

importance of sentences. Information Processing and Management, 40, 65–79.

https://doi.org/10.1016/S0306-4573(02)00056-0.

Kolb, P. (2009). Experiments on the difference between semantic similarity and

150

relatedness. Proceedings of Nodalida 2009 (pp. 81–88).

Kondrak, G. (2005). N -Gram Similarity and Distance. Lecture Notes in Computer

Science, 3772, 115–126. https://doi.org/10.1.1.69.2441.

Landauer, T. K., & Dumais, S. T. (1997). A solution to Plato’s problem: The latent

semantic analysis theory of acquisition, induction, and representation of

knowledge. Psychological Review. https://doi.org/10.1037/0033-

295X.104.2.211.

Lapata, M., & Barzilay, R. (2005). Automatic evaluation of text coherence: Models

and representations. IJCAI International Joint Conference on Artificial

Intelligence (pp. 1085–1090).

Leacock, C., & Chodorow, M. (1998). Combining Local Context and WordNet

Similarity for Word Sense Identification. WordNet: An electronic lexical

database. (pp. 265–283). https://doi.org/citeulike-article-id:1259480.

Lesk, M. (1986). Automatic sense disambiguation using machine readable

dictionaries. Proceedings of the 5th annual international conference on Systems

documentation - SIGDOC ’86 (pp. 24–26).

https://doi.org/10.1145/318723.318728.

Li, H., Tian, Y., & Cai, Q. (2011). Improvement of semantic similarity algorithm

based on WordNet. Proceedings of the 2011 6th IEEE Conference on Industrial

Electronics and Applications, ICIEA 2011 (pp. 564–567).

https://doi.org/10.1109/ICIEA.2011.5975649.

Li, Y., McLean, D., Bandar, Z. A., O’Shea, J. D., & Crockett, K. (2006). Sentence

similarity based on semantic nets and corpus statistics. IEEE Transactions on

Knowledge and Data Engineering, 18(8), 1138–1150.

https://doi.org/10.1109/TKDE.2006.130.

Lin, C.-Y., & Hovy, E. (2003). Automatic evaluation of summaries using N-gram co-

occurrence statistics. Proceedings of the 2003 Conference of the North

American Chapter of the Association for Computational Linguistics on Human

Language Technology - NAACL ’03, 2003, 71–78.

https://doi.org/10.3115/1073445.1073465.

Lin, D. (1998a). An Information-Theoretic Definition of Similarity. Proceedings of

ICML, 296–304.

Lin, D. (1998b). Extracting Collocations from Text Corpora. First Workshop on

Computational Terminology (pp. 57–63). https://doi.org/10.1.1.56.1687.

151

Liu, Y., & Zong, C. (2004). Example-based Chinese-english MT. Conference

Proceedings - IEEE International Conference on Systems, Man and Cybernetics

(Vol. 7, pp. 6093–6096). https://doi.org/10.1109/ICSMC.2004.1401354.

Liu, H., Bao, H., & Xu, D. (2012). Concept vector for semantic similarity and

relatedness based on WordNet structure. Journal of Systems and Software (Vol.

85, pp. 370–381). https://doi.org/10.1016/j.jss.2011.08.029

Lund, K., Burgess, C., & Atchley, R. A. (1995). Semantic and Associative Priming

in High-Dimensional Semantic Space. In Cognitive Science Proceedings, LEA

(pp. 660–665). https://doi.org/10.1016/j.jconhyd.2010.

08.009.

Madhavan, J., Bernstein, P. A., Doan, A., & Halevy, A. (2005). Corpus-based

schema matching. In Proceedings - International Conference on Data

Engineering (pp. 57–68). https://doi.org/10.1109/ICDE.2005.39.

Maind, A., Deorankar, A., & Chatur, P. (2012). Measurement Of Semantic Similarity

Between Words: A Survey. International Journal of Computer Science,

Engineering & Informa;, 2(6), p51.

Martinez-Gil, J., & Aldana-Montes, J. F. (2013). Semantic similarity measurement

using historical google search patterns. Information Systems Frontiers, 15(3),

399–410. https://doi.org/10.1007/s10796-012-9404-7.

Miller, G. A. (1995). WordNet: a lexical database for English. Communications of

the ACM, 38, 39–41. https://doi.org/10.1145/219717.219748.

Miller, G. A., & Charles, W. G. (1991). Contextual correlates of semantic similarity.

Language and Cognitive Processes. https://doi.org/10.1080/0169096910840

6936.

Mohameth-François., Ranwez, S., Montmain, J., Regnault, A., Crampes, M., &

Ranwez, V. (2012). User centered and ontology based information retrieval

system for life sciences. BMC Bioinformatics, 13 Suppl 1, S4.

https://doi.org/10.1186/1471-2105-13-S1-S4.

Needleman, S. B., & Wunsch, C. D. (1970). A general method applicable to the

search for similarities in the amino acid sequence of two proteins. Journal of

Molecular Biology, 48, 443–453. https://doi.org/10.1016/0022-2836(70)900574.

Neter, J., Kutner, MH., Nachtsheim,CJ., Wasserman, W., (1996). Applied Linear

Statistical Modelslrwin, Chicago.

Noy, N., & McGuinness, D. (2001). Ontology development 101: A guide to creating

152

your first ontology. Development, 32, 1–25.

https://doi.org/10.1016/j.artmed.2004.01.014.

Ochiai A. (1957). Zoogeographical studies on the soleoid fishes found Japan and its

neighboring regions. Japan Society Fish Sciences, 22:526-30.

Palmer, M., & Wu, Z. (1994). Verb Semantics And Lexical. Proceeding ACL ’94

Proceedings of the 32nd Annual Meeting on Association for Computational

Linguistics, 133–138.

Papineni, K., Roukos, S., Ward, T., & Zhu, W. (2002). BLEU: a method for

automatic evaluation of machine translation. Proceedings of the 40th annual

meeting on association for computational linguistics, 311–318.

https://doi.org/10.3115/1073083.1073135

Patwardhan, S., & Pedersen, T. (2006). Using WordNet-based Context Vectors to

Estimate the Semantic Relatedness of Concepts. 11th Conference of the

European Chapter of the Association for Computational Linguistics, 1501, 1–8.

https://doi.org/citeulike-article-id:1574418.

Pedersen, T., Pakhomov, S. V. S., Patwardhan, S., & Chute, C. G. (2007). Measures

of semantic similarity and relatedness in the biomedical domain. Journal of

Biomedical Informatics, 40, 288–299. https://doi.org/10.1016/j.jbi.2006.06.004.

Pesquita, C., Faria, D., Falcão, A. O., Lord, P., & Couto, F. M. (2009). Semantic

similarity in biomedical ontologies. PLoS Computational Biology.

https://doi.org/10.1371/journal.pcbi.1000443.

Petrakis, E., Varelas, G., Hliaoutakis, A., & Raftopoulou, P. (2006). X-Similarity:

Computing Semantic Similarity between Concepts from Different Ontologies.

Journal of Digital Information Management, 4(4), 233. Retrieved from

http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.114.3247.

Pirró, G. (2009). A semantic similarity metric combining features and intrinsic

information content. Data and Knowledge Engineering, 68, 1289–1308.

https://doi.org/10.1016/j.datak.2009.06.008.

Pirró, G., & Euzenat, J. (2010). A feature and information theoretic framework for

semantic similarity and relatedness. The Semantic Web–ISWC 2010. Retrieved

from http://link.springer.com/chapter/10.1007/978-3-642-17746-0_39.

Pirrò, G., Ruffolo, M., & Talia, D. (2009). SECCO: On building semantic links in

peer-to-peer networks. Lecture Notes in Computer Science (including subseries

Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

153

(Vol. 5480 LNCS, pp. 1–36). https://doi.org/10.1007/978-3-642-00685-2_1.

Pirró, G., & Seco, N. (2008). Design , Implementation and Evaluation of a New

Semantic Similarity Metric Combining Features and Intrinsic Information

Content. Lecture Notes in Computer Science Volume 5332 (Vol. 5332, pp.

1271–1288). https://doi.org/10.1007/978-3-540-88873-4.

Quine, W. V. O. (1969). Ontological Relativity & Other Essays. Ontological

Relativity & Other Essays, 1–25. https://doi.org/10.2307/2024305.

Rada, R., Mili, H., Bicknell, E., & Blettner, M. (1989). Development and application

of a metric on semantic nets. IEEE Transactions on Systems, Man, and

Cybernetics, 19(1), 17–30. https://doi.org/10.1109/21.24528.

Resnik, P. (1995). Using information content to evaluate seantic similarity in a

taxonomy. Proceedings of the 14th International Joint Conference on Artificial

Intelligence (IJCAI).

Richardson, R., Smeaton, A., & Murphy, J. (1994). Using WordNet as a knowledge

base for measuring semantic similarity between words. Retrieved from

http://ducra.dcu.ie/wpapers/1994/1294.pdf.

Rodríguez, M. A., & Egenhofer, M. J. (2003a). Comparing Geospatial Entity

Classes : An Asymmetric and Context-Dependent Similarity Measure.

International Journal of Geographical Information Science, vol. 18, no. 3, pp.

229–256, 2004

Rodríguez, M. A., & Egenhofer, M. J. (2003b). Determining semantic similarity

among entity classes from different ontologies. IEEE Transactions on

Knowledge and Data Engineering, 15, 442–456. https://doi.org/10.1109/TKDE

.2003.1185844.

Rubenstein, H., & Goodenough, J. B. (1965). Contextual correlates of synonymy.

Communications of the ACM. https://doi.org/10.1145/365628.365657.

Saleena, B., & Srivatsa, S. K. (2014). Using concept similarity in cross ontology for

adaptive e-Learning systems. Journal of King Saud University - Computer and

Information Sciences, 27(1), 1–12. https://doi.org/10.1016/j.jksuci.2014.03.007.

Sánchez, D., & Batet, M. (2012). A new model to compute the information content

of concepts from taxonomic knowledge. International Journal on Semantic Web

and Information Systems, 8, 34–50. https://doi.org/10.4018/jswis.2012040102.

Sánchez, D., & Batet, M. (2011). Semantic similarity estimation in the biomedical

domain: An ontology-based information-theoretic perspective. Journal of

154

Biomedical Informatics, 44, 749–759. https://doi.org/10.1016/j.jbi.2011.03.013.

Sánchez, D., & Batet, M. (2013). A semantic similarity method based on information

content exploiting multiple ontologies. Expert Systems with Applications, 40(4),

1393–1399. https://doi.org/10.1016/j.eswa.2012.08.049.

Sánchez, D., Batet, M., Isern, D., & Valls, A. (2012). Ontology-based semantic

similarity: A new feature-based approach. Expert Systems with Applications,

39(9), 7718–7728. https://doi.org/10.1016/j.eswa.2012.01.082.

Sánchez, D., Batet, M., Valls, A., & Gibert, K. (2010). Ontology-driven web-based

semantic similarity. Journal of Intelligent Information Systems, 35, 383–413.

https://doi.org/10.1007/s10844-009-0103-x.

Sánchez, D., & Isern, D. (2011). Automatic extraction of acronym definitions from

the Web. Applied Intelligence, 34, 311–327. https://doi.org/10.1007/s10489-

009-0197-4.

Sánchez, D., Isern, D., & Millan, M. (2011). Content annotation for the semantic

web: An automatic web-based approach. Knowledge and Information Systems,

27, 393–418. https://doi.org/10.1007/s10115-010-0302-3.

Sánchez, D., Solé-Ribalta, A., Batet, M., & Serratosa, F. (2012). Enabling semantic

similarity estimation across multiple ontologies: an evaluation in the biomedical

domain. Journal of Biomedical Informatics, 45(1), 141–55.

https://doi.org/10.1016/j.jbi.2011.10.005.

Saruladha, K., Aghila, G., & Bhuvaneswary, A. (2011a). COSS: Cross Ontology

Semantic Similarity measure-An information content based approach. 2011

International Conference on Recent Trends in Information Technology

(ICRTIT), 485–490. https://doi.org/10.1109/ICRTIT.2011.5972360.

Saruladha, K., Aghila, G., & Bhuvaneswary, A. (2011b). Information content based

semantic similarity approaches for multiple biomedical ontologies.

Communications in Computer and Information Science (Vol. 191 CCIS, pp.

327–336). https://doi.org/10.1007/978-3-642-22714-1_34.

Schallehn, E., Sattler, K. U., & Saake, G. (2004). Efficient similarity-based

operations for data integration. Data and Knowledge Engineering, 48, 361–387.

https://doi.org/10.1016/j.datak.2003.08.004.

Schickel-Zuber, V., & Faltings, B. (2007). OSS: A semantic similarity function

based on hierarchical ontologies. IJCAI International Joint Conference on

Artificial Intelligence (pp. 551–556).

155

Schütze, H. (1998). Automatic Word Sense Discrimination. Computational

Linguistics, 24, 97–123.

Schwartzman, S. (1996). The Words of Mathematics: An Etymological Dictionary of

Mathematical Terms Used in English (Spectrum). The Mathematical

Association of America (September 1996).

Schwering, A. (2008). Approaches to semantic similarity measurement for geo-

spatial data: A survey. Transactions in GIS. https://doi.org/10.1111/j.1467-

9671.2008.01084.x.

Singh, P. (2004). Web Ontology to Facilitate Semantic Web. 2nd International

CALIBER-2004, New Delhi, 11-13 February, 2004 (pp. 11–13).

Slimani, T. (2013). Description and Evaluation of Semantic similarity Measures

Approaches. Journal of Computer Applicationsof Computer Applications, 80–

no 10, 1–10.

Smith, T. F., & Waterman, M. S. (1981). Identification of common molecular

subsequences. Journal of Molecular Biology, 147, 195–197.

https://doi.org/10.1016/0022-2836(81)90087-5.

Solé-Ribalta, A., Sánchez, D., Batet, M., & Serratosa, F. (2014). Towards the

estimation of feature-based semantic similarity using multiple ontologies.

Knowledge-Based Systems, 55, 101–113. https://doi.org/10.1016/j.knosys.

2013.10.015.

Spasic, I., Ananiadou, S., McNaught, J., & Kumar, A. (2005). Text mining and

ontologies in biomedicine: Making sense of raw text. Briefings in

Bioinformatics, 6, 239–251. https://doi.org/10.1093/bib/6.3.239.

Studer, R., Benjamins, V. R., & Fensel, D. (1998). Knowledge engineering:

Principles and methods. Data & Knowledge Engineering, 25(1–2), 161–197.

https://doi.org/10.1016/S0169-023X(97)00056-6.

Sussna, M. (1993). Word sense disambiguation using a massive of computer for free-

text semantic indexing network. CIKM ’93 Proceedings of the Second

International Conference on Information and Knowledge Management, 67–74.

https://doi.org/10.1145/170088.170106.

Thiagarajan, R., Manjunath, G., & Stumptner, M. (2008). Computing Semantic

Similarity Using Ontologies. Computing, Germany. Retrieved from

http://www.hpl.hp.com/techreports/2008/HPL-2008-87.pdf

156

Tversky, A. (1977). Features of similarity. Psychological Review.

https://doi.org/10.1037/0033-295X.84.4.327.

Vicient, C., Sánchez, D., & Moreno, A. (2013). An automatic approach for ontology-

based feature extraction from heterogeneous textual resources. Engineering

Applications of Artificial Intelligence, 26, 1092–1106.

https://doi.org/10.1016/j.engappai.2012.08.002.

Wache, H., Vögele, T., Visser, U., Stuckenschmidt, H., Schuster, G., Neumann, H.,

& Hübner, S. (2001). Ontology-Based Integration of Information - A Survey of

Existing Approaches. IJCAI-01 Workshop: Ontologies and Information Sharing

(pp. 108–117). https://doi.org/10.1016/j.ijhcs.2008.07.007.

Wang, L., & Chen, M. (2013). A New Model to Compute Semantic Similarity from

Multi-ontology. China Semantic Web Symposium and Web Science Conference.

Communications in Computer and Information Science book series (CCIS, vol:

406) 1–10.

Wang, J. Z., Du, Z., Payattakool, R., Yu, P. S., & Chen, C.-F. (2007). A new method

to measure the semantic similarity of GO terms. Bioinformatics (Oxford,

England), 23, 1274–1281. https://doi.org/10.1093/bioinformatics/btm087.

Wegrzyn-Wolska, K., & Szczepaniak, P. S. (2005). Classification of RSS-Formatted

Documents Using Full Text Similarity Measures. Lecture Notes in Computer

Science, 3579, 400–405.

Winkler, W. E. (1990). String Comparator Metrics and Enhanced Decision Rules in

the Fellegi-Sunter Model of Record Linkage. Proceedings of the Section on

Survey Research (pp. 354–359).

Yatskevich, M., & Giunchiglia, F. (2007). Element level semantic matching using

WordNet. CiteSeerX.

Yuhua Li, Zuhair A. Bandar, and D. M. (2003). An approach for measuring

semantic similarity between words using multiple information sources. IEEE

Transactions on Knowledge and Data Engineering, 15(4), 871–882.

https://doi.org/10.1109/TKDE.2003.1209005.

Zhang, S., Zheng, X., & Hu, C. (2015). A Survey of Semantic Similarity and its

Application to Social Network Analysis. IEEE International Conference on Big

Data (Big Data) (pp. 1–3).

https://link.springer.com/conference/csemws

https://link.springer.com/bookseries/7899